@mgaudet@discuss.systems @shriramk@mastodon.social @cfbolz@mastodon.social @shu@mstdn.social @jandem@mastodon.social If we're settling for reminiscences: To expand on what @shu said, the benchmark by which we measured the success of performance efforts at the time was SunSpider, which IIRC was a collection of C benchmarks ported to JS - it just was nothing like the real workload. The lesson I take away is not "don't do tracing" so much as "don't chase bad benchmarks"